2 * Copyright (c) 1997, 1998
3 * Nan Yang Computer Services Limited. All rights reserved.
5 * This software is distributed under the so-called ``Berkeley
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by Nan Yang Computer
20 * 4. Neither the name of the Company nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * This software is provided ``as is'', and any express or implied
25 * warranties, including, but not limited to, the implied warranties of
26 * merchantability and fitness for a particular purpose are disclaimed.
27 * In no event shall the company or contributors be liable for any
28 * direct, indirect, incidental, special, exemplary, or consequential
29 * damages (including, but not limited to, procurement of substitute
30 * goods or services; loss of use, data, or profits; or business
31 * interruption) however caused and on any theory of liability, whether
32 * in contract, strict liability, or tort (including negligence or
33 * otherwise) arising in any way out of the use of this software, even if
34 * advised of the possibility of such damage.
36 * $Id: request.h,v 1.21 2001/05/23 23:03:23 grog Exp grog $
40 /* Information needed to set up a transfer */
44 XFR_NORMAL_WRITE = 2, /* write request in normal mode */
45 XFR_RECOVERY_READ = 4,
46 XFR_DEGRADED_WRITE = 8,
47 XFR_PARITYLESS_WRITE = 0x10,
48 XFR_NO_PARITY_STRIPE = 0x20, /* parity stripe is not available */
49 XFR_DATA_BLOCK = 0x40, /* data block in request */
50 XFR_PARITY_BLOCK = 0x80, /* parity block in request */
51 XFR_BAD_SUBDISK = 0x100, /* this subdisk is dead */
52 XFR_MALLOCED = 0x200, /* this buffer is malloced */
54 XFR_PHASE2 = 0x800, /* documentation only: 2nd phase write */
56 XFR_REVIVECONFLICT = 0x1000, /* possible conflict with a revive operation */
57 XFR_BUFLOCKED = 0x2000, /* BUF_LOCK performed on this buffer */
58 XFR_COPYBUF = 0x4000, /* data buffer was copied */
59 /* operations that need a parity block */
60 XFR_PARITYOP = (XFR_NORMAL_WRITE | XFR_RECOVERY_READ | XFR_DEGRADED_WRITE),
61 /* operations that use the group parameters */
62 XFR_GROUPOP = (XFR_DEGRADED_WRITE | XFR_RECOVERY_READ),
63 /* operations that that use the data parameters */
64 XFR_DATAOP = (XFR_NORMAL_READ | XFR_NORMAL_WRITE | XFR_PARITYLESS_WRITE),
65 /* operations requiring read before write */
66 XFR_RBW = (XFR_NORMAL_WRITE | XFR_DEGRADED_WRITE),
67 /* operations that need a malloced buffer */
68 XFR_NEEDS_MALLOC = (XFR_NORMAL_WRITE | XFR_RECOVERY_READ | XFR_DEGRADED_WRITE)
72 * Describe one low-level request, part of a
73 * high-level request. This is an extended
74 * struct buf buffer, and the first element
75 * *must* be a struct buf. We pass this
76 * structure to the I/O routines instead of a
77 * struct buf in order to be able to locate the
78 * high-level request when it completes.
80 * All offsets and lengths are in sectors.
84 struct buf b; /* buf structure */
85 struct rqgroup *rqg; /* pointer to our group */
86 /* Information about the transfer */
87 daddr_t sdoffset; /* offset in subdisk */
88 int useroffset; /* offset in user buffer of normal data */
90 * dataoffset and datalen refer to "individual" data
91 * transfers which involve only this drive (normal read,
92 * parityless write) and also degraded write.
94 * groupoffset and grouplen refer to the other "group"
95 * operations (normal write, recovery read) which involve
96 * more than one drive. Both the offsets are relative to
97 * the start of the local buffer.
99 int dataoffset; /* offset in buffer of the normal data */
100 int groupoffset; /* offset in buffer of group data */
101 short datalen; /* length of normal data (sectors) */
102 short grouplen; /* length of group data (sectors) */
103 short buflen; /* total buffer length to allocate */
104 short flags; /* really enum xferinfo (see above) */
105 /* Ways to find other components */
106 short sdno; /* subdisk number */
107 short driveno; /* drive number */
111 * A group of requests built to satisfy an I/O
112 * transfer on a single plex.
115 struct rqgroup *next; /* pointer to next group */
116 struct request *rq; /* pointer to the request */
117 short count; /* number of requests in this group */
118 short active; /* and number active */
119 short plexno; /* index of plex */
120 int badsdno; /* index of bad subdisk or -1 */
121 enum xferinfo flags; /* description of transfer */
122 struct rangelock *lock; /* lock for this transfer */
123 daddr_t lockbase; /* and lock address */
124 struct rqelement rqe[0]; /* and the elements of this request */
128 * Describe one high-level request and the
129 * work we have to do to satisfy it.
132 struct buf *bp; /* pointer to the high-level request */
133 caddr_t save_data; /* for copied write buffers */
136 int volno; /* volume index */
137 int plexno; /* or plex index */
139 int error; /* current error indication */
140 int sdno; /* reviving subdisk (XFR_REVIVECONFLICT) */
141 short isplex; /* set if this is a plex request */
142 short active; /* number of subrequests still active */
143 struct rqgroup *rqg; /* pointer to the first group of requests */
144 struct rqgroup *lrqg; /* and to the last group of requests */
145 struct request *next; /* link of waiting requests */
149 * Extended buffer header for subdisk I/O. Includes
150 * a pointer to the user I/O request.
153 struct buf b; /* our buffer */
154 struct buf *bp; /* and pointer to parent */
155 short driveno; /* drive index */
156 short sdno; /* and subdisk index */
160 * Values returned by rqe and friends. Be careful
161 * with these: they are in order of increasing
162 * seriousness. Some routines check for
163 * > REQUEST_RECOVERED to indicate a failed request. XXX
166 REQUEST_OK, /* request built OK */
167 REQUEST_RECOVERED, /* request OK, but involves RAID5 recovery */
168 REQUEST_DEGRADED, /* parts of request failed */
169 REQUEST_EOF, /* parts of request failed: outside plex */
170 REQUEST_DOWN, /* all of request failed: subdisk(s) down */
171 REQUEST_ENOMEM /* all of request failed: ran out of memory */
175 /* Trace entry for request info (DEBUG_LASTREQS) */
177 loginfo_unused, /* never been used */
178 loginfo_user_bp, /* this is the bp when strategy is called */
179 loginfo_user_bpl, /* and this is the bp at launch time */
180 loginfo_rqe, /* user RQE */
181 loginfo_iodone, /* iodone */
182 loginfo_raid5_data, /* write RAID-5 data block */
183 loginfo_raid5_parity, /* write RAID-5 parity block */
184 loginfo_sdio, /* subdisk I/O */
185 loginfo_sdiol, /* subdisk I/O launch */
186 loginfo_sdiodone, /* subdisk iodone */
187 loginfo_lockwait, /* wait for range lock */
188 loginfo_lock, /* lock range */
189 loginfo_unlock, /* unlock range */
193 * This is the rangelock structure with an added
194 * buffer pointer and plex number. We don't need
195 * the plex number for the locking protocol, but
196 * it does help a lot when logging.
198 struct rangelockinfo {
199 daddr_t stripe; /* address + 1 of the range being locked */
200 struct buf *bp; /* user's buffer pointer */
204 union rqinfou { /* info to pass to logrq */
206 struct rqelement *rqe; /* address of request, for correlation */
207 struct rangelockinfo *lockinfo;
211 enum rqinfo_type type; /* kind of event */
212 struct timeval timestamp; /* time it happened */
213 struct buf *bp; /* point to user buffer */
214 int devmajor; /* major and minor device info */
217 struct buf b; /* yup, the *whole* buffer header */
218 struct rqelement rqe; /* and the whole rqe */
219 struct rangelock lockinfo;
223 #define RQINFO_SIZE 128 /* number of info slots in buffer */
225 void logrq(enum rqinfo_type type, union rqinfou info, struct buf *ubp);
228 /* Structures for the daemon */
230 /* types of request to the daemon */
232 daemonrq_none, /* dummy to catch bugs */
233 daemonrq_ioerror, /* error occurred on I/O */
234 daemonrq_saveconfig, /* save configuration */
235 daemonrq_return, /* return to userland */
236 daemonrq_ping, /* show sign of life */
237 daemonrq_init, /* initialize a plex */
238 daemonrq_revive, /* revive a subdisk */
239 daemonrq_closedrive, /* close a drive */
242 /* info field for daemon requests */
243 union daemoninfo { /* and the request information */
244 struct request *rq; /* for daemonrq_ioerror */
245 struct sd *sd; /* for daemonrq_revive */
246 struct plex *plex; /* for daemonrq_init */
247 struct drive *drive; /* for daemonrq_closedrive */
248 int nothing; /* for passing NULL */
252 struct daemonq *next; /* pointer to next element in queue */
253 enum daemonrq type; /* type of request */
254 int privateinuse; /* private element, being used */
255 union daemoninfo info; /* and the request information */
258 void queue_daemon_request(enum daemonrq type, union daemoninfo info);
260 extern int daemon_options;
263 daemon_verbose = 1, /* talk about what we're doing */
265 daemon_noupdate = 4, /* don't update the disk config, for recovery */
268 void freerq(struct request *rq);
269 void unlockrange(int plexno, struct rangelock *);
270 /* Local Variables: */
271 /* fill-column: 50 */